Dear Dynare users,
I have noticed an inconsistency when analyzing smoothed variables produced using posterior mode vs. posterior mean. When the observable data has a non-zero steady state (the data is not demeaned prior to estimation, and I set the steady states in .mod file equal to sample means), using estimated posterior mode I get in oo_.SmoothedVariables.obs_data a smoothed series which I believe is represented in deviations form steady state as it has a zero mean. But when doing some MH draws and using ‘smoother’ in estimation command, the same smoothed observable variable, located now in oo_.SmoothedVariables.Mean.obs_data, is identical to the actual observable, i.e. is not in deviations from steady state.
If anybody encountered the same issue, please confirm.
Also, I believe the treatment should be the same, i.e. smoothed variables constructed under the two alternatives should be both stored as deviations from steady state, or as actual levels.
I am using Dynare 4.4.2.